[lammps-users] reaxFF f/c performance


I made a very small test on the performance of reaxFF (56 atoms during 1000 steps, NVT) in the fortran and C implementations. They gave exactly the same numbers (I create the velocities with the same seed), and the walltime was basically equal. Is there any reason why we should prefer one over the other? Should it be the new C implementation faster for big systems?


Try a bigger problem to see a speed difference.
See the benchmark page on the WWW site - last
section on potentials for some numbers.